ci-linux-devel — Discussion about Cluster Infrastructure for Linux

[CI] Query

[Ramachandra P. <lin...@ya...>]

Hi,
    Trying to educate myself on CI. 

    Looking at the features published on the web page,
I did not find any support for global reliable ordered
messaging between nodes of the cluster. 

    How does any subsystem relying on CI make sure
that messages sent by any node in the cluster to
everybody else is seen in the same order on all the
nodes. Isn't this kind of feature neccessary for SSI
or any other clients of CI? If so how is that
accomplished?


Thanks,
Ram Pai 

__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail
http://personal.mail.yahoo.com/

[CI] CI 0.5.5 is coming

[Brian J. W. <Bri...@co...>]

We'll release version 0.5.5 tommorrow or Wednesday. The primary change
is in how a node joins the cluster. Instead of passing the node
number, IP addrs, etc. through LILO, they'll be passed through a
system call sometime after mounting the root and starting process 1.
The system call itself will be undocumented, but we have a command
that reads a configuration file and makes the system call. We'll clean
this up a bit more in the future. ;)

How this would work for an SSI cluster is that the node would mount a
local ramdisk and start a modified linuxrc as process 1. In addition
to its usual stuff, linuxrc would run our command to join the cluster,
then it would pivot_root() the clusterwide shared root into place.
Finally, it would exec init if it's the first node in the cluster, or
exit otherwise. (Please correct me if I'm wrong, Dave)

For more loosely coupled clusters, the node could join at any time.
Updated documentation on how to do this will be provided.


Aneesh-

Can you adapt your Alpha patch to the new release? We'll then
double-check that it still works on Intel x86, and fold it into the
subsequent release.

Thanks for your efforts.


-- 
Brian Watson            |  "The common people of England... so jealous
Linux Kernel Developer  |  of their liberty, but like the common
Compaq Computer Corp    |  people of most other countries never
Los Angeles, CA         |  rightly considering wherein it consists..."
310-414-3119            |      -Adam Smith, Wealth of Nations, 1776

Re: [CI] Who are you?

[Bruce W. <br...@ka...>]

Who are we?
   Going back to the 80's, we developed a technology
called the Locus Distributed Systems, which was a full
SSI cluster technology that was both BSD and SysV.  We
were contracted by IBM to build TCF (Transparent Computing
Facility) that was a product on the PS/2 and 370
around 1989.  Starting in about 1991, we began
rewriting all the SSI code, first on OSF/1-AD as
part of the Intel Paragon offering, and then on
SysV4.  We  (Locus computing), then contracted
with Tandem in about 1992 to build SSI HA
clusters for the S4000 (hardware fault tolerance
Unix offering Tandem had primarily for Telcos).  The
technology and a group of the engineers/architects
was aquired by Tandem in 1996 so we could bring out
a version on Intel, using Unixware as the host.
That product is called NonStop Clusters for Unixware.
We started porting the code from Unixware to Linux
about a year ago.  CI is the first 2 pieces that
we believe can be the framework for not only SSI
clusters but also HPTC clusters and other forms of
Linux clusters.

I very much appreciate your interest and would love
to have your help.  We are not doing a "product", 
however.  This is a technolgy offering that we hope
will be embrassed by the open source community and
eventually be part of distributions and products.

bruce

[Charset iso-8859-1 unsupported, filtering to ASCII...]
> 	
> 
> 	Which is a question I was asked this week.
> 
> 	I am a HPTC consultant involved in both high performance Linux
> clusters as well as Tru64 high availability clusters. In short, I'd like to
> talk to you about what you are doing and how we can help your group to get
> your product to our customers.
> 
> 	Call me at (972) 679-7892 please...
> 
> 	Henry Schwindt
> _______________________________________________
> ci-linux-devel mailing list
> ci-...@op...
> http://www.opensource.compaq.com/mailman/listinfo/ci-linux-devel

[CI] Who are you?

[Schwindt, H. <Hen...@co...>]

	

	Which is a question I was asked this week.

	I am a HPTC consultant involved in both high performance Linux
clusters as well as Tru64 high availability clusters. In short, I'd like to
talk to you about what you are doing and how we can help your group to get
your product to our customers.

	Call me at (972) 679-7892 please...

	Henry Schwindt

[CI] [Fwd: [PATCH] Cluster Infrastructure for Linux -Alpha Platform (SSI cluster)]

[Brian J. W. <Bri...@co...>]

Hi ,

 Forgot to send  this patch.

 -aneesh

[CI] [PATCH]Cluster Tools patch to make it work on Alpha Linux

[Brian J. W. <Bri...@co...>]

Hi ,

 Some minor changes needed to make cluster tools work on Alpha  Linux.

 Note: I am not able to send mails to project mailing list because of
the exchange server issues over here. We are trying to fix it.

-aneesh

[CI] Sending It again Cluster tools patch (Alpha Linux )

[<kva...@ya...>]

diff -Nru
cluster-tools-0.5.0/cmd/clusternode_getstate.c
cluster-tools-0.5.0-alpha/cmd/clusternode_getstate.c
--- cluster-tools-0.5.0/cmd/clusternode_getstate.c    
 Thu Jun  7 05:34:50 2001
+++
cluster-tools-0.5.0-alpha/cmd/clusternode_getstate.c  
     Fri Jun 22 15:20:01 2001
@@ -16,6 +16,14 @@
  *     Questions/Comments/Bugfixes to
ci-...@li...
  *
  */
+/*
+ *
+ * 22JUN2001 ane...@di...
+ * We need to check return values after calling
function  clusternode_info.
+ * Otherwise following 'printf' will dump 
+ * 
+ */
+
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -71,8 +79,13 @@
                exit(1);
        }
 
-       (void)clusternode_info(node, sizeof(ni), &ni);
-       printf("%s\n", states[ni.node_state]);
+       if(clusternode_info(node, sizeof(ni), &ni) >=
0) {
+               printf("%s\n", states[ni.node_state]);
+       }
+       else {
+               printf("Error in getting the cluster
information \n");
+       }
+       
        
        exit(0);
 }
diff -Nru
cluster-tools-0.5.0/libcluster/clusternode_info.c
cluster-tools-0.5.0-alpha/libcluster/clusternode_info.c
--- cluster-tools-0.5.0/libcluster/clusternode_info.c 
 Thu Jun  7 05:34:50 2001
+++
cluster-tools-0.5.0-alpha/libcluster/clusternode_info.c
    Fri Jun 15 10:36:19 2001
@@ -35,6 +35,8 @@
        iovec.tio_udatain = (caddr_t)&args;
        iovec.tio_udatainlen = sizeof(args);
        iovec.tio_udataout = (caddr_t)nodeinfo;
+/* outlen is check inside kernel to be != 0 
ane...@di... */
+       iovec.tio_udataoutlen =
sizeof(clusternode_info_t);
        
        return ssisys((char *)&iovec, sizeof(iovec));
 }


____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

[CI] Sending it again PATCH Alpha Linux CI-Linux

[<kva...@ya...>]

diff -Nru linux-2.4.2/arch/alpha/config.in
linux/arch/alpha/config.in
--- linux-2.4.2/arch/alpha/config.in    Sat Dec 30
03:37:19 2000
+++ linux/arch/alpha/config.in  Fri Jun 22 10:54:10
2001
@@ -363,3 +363,17 @@
 bool 'Legacy kernel start address'
CONFIG_ALPHA_LEGACY_START_ADDRESS
 
 endmenu
+
+mainmenu_option next_comment
+comment 'Clustering'
+bool 'Clustering' CONFIG_CLUSTER 
+if [ "$CONFIG_CLUSTER" = "y" ]; then
+  bool 'CLuster Membership Service (CLMS)'
CONFIG_CLMS
+  if [ "$CONFIG_CLMS" = "y" ]; then
+    comment 'Internode Communication Subsystem (ICS)
is required for CLMS'
+    define_bool CONFIG_ICS y
+  fi
+fi
+
+endmenu
+
diff -Nru linux-2.4.2/arch/alpha/defconfig
linux/arch/alpha/defconfig
--- linux-2.4.2/arch/alpha/defconfig    Tue Oct 17
04:08:41 2000
+++ linux/arch/alpha/defconfig  Fri Jun 22 10:54:42
2001
@@ -637,3 +637,10 @@
 CONFIG_MATHEMU=y
 CONFIG_MAGIC_SYSRQ=y
 CONFIG_ALPHA_LEGACY_START_ADDRESS=y
+
+#
+# Clustering 
+#
+CONFIG_CLUSTER=y
+CONFIG_CLMS=y
+CONFIG_ICS=y
diff -Nru linux-2.4.2/arch/alpha/kernel/entry.S
linux/arch/alpha/kernel/entry.S
--- linux-2.4.2/arch/alpha/kernel/entry.S       Mon
Sep  4 00:06:45 2000
+++ linux/arch/alpha/kernel/entry.S     Fri Jun 22
11:50:38 2001
@@ -2,6 +2,9 @@
  * alpha/entry.S
  *
  * kernel entry-points
+ *
+ * Incremented NR_SYSCALLS to include SSI system
calls 
+ * (ane...@di...)
  */
 
 #include <linux/config.h>
@@ -10,7 +13,7 @@
 
 #define SIGCHLD 20
 
-#define NR_SYSCALLS 377
+#define NR_SYSCALLS 379
 
 /*
  * These offsets must match with alpha_mv in
<asm/machvec.h>.
@@ -1022,7 +1025,7 @@
        .quad alpha_ni_syscall
        .quad alpha_ni_syscall
        .quad alpha_ni_syscall
-       .quad alpha_ni_syscall                  /* 240
*/
+       .quad alpha_ni_syscall                  /* 240
*/
        .quad osf_sysinfo
        .quad alpha_ni_syscall
        .quad alpha_ni_syscall
@@ -1161,3 +1164,5 @@
        .quad sys_mincore                       /* 375
*/
        .quad sys_pciconfig_iobase
        .quad sys_getdents64
+/* BEGIN SSI SYSTEM CALLS HERE
ane...@di... */
+       .quad sys_ssisys                      /* 378
*/
diff -Nru linux-2.4.2/arch/alpha/kernel/signal.c
linux/arch/alpha/kernel/signal.c
--- linux-2.4.2/arch/alpha/kernel/signal.c      Thu
Jan 25 04:46:23 2001
+++ linux/arch/alpha/kernel/signal.c    Fri Jun 22
10:59:20 2001
@@ -4,6 +4,8 @@
  *  Copyright (C) 1995  Linus Torvalds
  *
  *  1997-11-02  Modified for POSIX.1b signals by
Richard Henderson
+ * 
+ * Added SIGCLUSTER to do_signal() function
(ane...@di...)
  */
 
 #include <linux/sched.h>
@@ -690,6 +692,7 @@
 
                        switch (signr) {
                        case SIGCONT: case SIGCHLD:
case SIGWINCH:
+                       case SIGCLUSTER:
                                continue;
 
                        case SIGTSTP: case SIGTTIN:
case SIGTTOU:
diff -Nru linux-2.4.2/cluster/clms/clms_ics.c
linux/cluster/clms/clms_ics.c
--- linux-2.4.2/cluster/clms/clms_ics.c Fri Jun 22
10:43:54 2001
+++ linux/cluster/clms/clms_ics.c       Fri Jun 22
11:04:14 2001
@@ -240,10 +240,18 @@
 /*
  * param OUT clock_t *
  */
+#ifdef __alpha__
+#define cli_decode_clock_t_p(hndl,x) \
+       cli_decode_as_llong_p(hndl,x,long)
+#define svr_encode_clock_t_p(hndl,x) \
+       svr_encode_as_llong_p(hndl,x,long)
+#else
 #define cli_decode_clock_t_p(hndl,x) \
        cli_decode_as_int_p(hndl,x,long)
 #define svr_encode_clock_t_p(hndl,x) \
        svr_encode_as_int_p(hndl,x,long)
+
+#endif 
 
 /*
  * Include prototypes, client and server stubs
generated by icsgen.
diff -Nru linux-2.4.2/cluster/clms/clms_mgmt.c
linux/cluster/clms/clms_mgmt.c
--- linux-2.4.2/cluster/clms/clms_mgmt.c        Fri
Jun 22 10:43:54 2001
+++ linux/cluster/clms/clms_mgmt.c      Fri Jun 22
11:08:06 2001
@@ -1518,11 +1518,14 @@
                return -EINVAL;  /* invalid node
number */
 
        if (node != CLUSTERNODE_THIS) {
-               if (node == this_node)
+               if (node == this_node) {
+       /* Just to correct the compiler warning
ane...@di... */
+                       printk("   ");
 #ifdef CONFIG_KDB
                        KDB_ENTER();
 #endif /* CONFIG_KDB */
-               else {
+               }
+               else 
                if (node != this_node) {
                        state =
clms_get_node_status(node);
                        if (!(state & CLMS_NODE_UP))
diff -Nru linux-2.4.2/cluster/ics/ics_tcp/ics_llcli.c
linux/cluster/ics/ics_tcp/ics_llcli.c
--- linux-2.4.2/cluster/ics/ics_tcp/ics_llcli.c Fri
Jun 22 10:43:55 2001
+++ linux/cluster/ics/ics_tcp/ics_llcli.c       Fri
Jun 22 11:16:22 2001
@@ -398,7 +398,8 @@
        int             nelem)
 {
        /* Assert word alignment */
-       ASSERT((u_int)param % sizeof(int64) == 0);
+/* param is an address ane...@di... */
+       ASSERT((u_long)param % sizeof(int64) == 0);
 
        ASSERT((void *)handle->ch_msg_inline_data_p +
(nelem<<INT64SHIFT) <=
               handle->ch_msg_inline_iov_p[0].iov_base
+
@@ -467,7 +468,8 @@
        int             nelem)
 {
        /* Assert word alignment */
-       ASSERT((u_int)param % sizeof(int32) == 0);
+/*param is an address ane...@di... */
+       ASSERT((u_long)param % sizeof(int32) == 0);
 
        ASSERT((void *)handle->ch_msg_inline_data_p +
(nelem<<INT32SHIFT) <=
               handle->ch_msg_inline_iov_p[0].iov_base
+
diff -Nru linux-2.4.2/cluster/ics/ics_tcp/ics_llmgmt.c
linux/cluster/ics/ics_tcp/ics_llmgmt.c
--- linux-2.4.2/cluster/ics/ics_tcp/ics_llmgmt.c      
 Fri Jun 22 10:43:55 2001
+++ linux/cluster/ics/ics_tcp/ics_llmgmt.c      Fri
Jun 22 11:20:08 2001
@@ -1006,7 +1006,7 @@
                if
(!clms_is_secondary_by_addr(&icsinfo)) {
                        printk(KERN_WARNING
                              "ics_llunack_daemon:
received request from unauthorized node (0x%x)",
-                             (unsigned int)
&icsinfo);
+                             (unsigned long)
&icsinfo);
                        continue;
                }
                /*
@@ -1317,7 +1317,7 @@
 ics_probe_clms_daemon(
        void                    *port_argp)
 {
-       int                     port;
+        long                   port;
        int                     error;
        struct sockaddr_in      sin;
        struct msghdr           *msg;
@@ -1333,7 +1333,7 @@
        int                     len;
 
 
-       port = (int) port_argp;
+       port = (long) port_argp;
 
        error = sock_create(PF_INET, SOCK_STREAM,
IPPROTO_TCP, &sock);
        if (error < 0)
@@ -1521,14 +1521,14 @@
 ics_chan_listen(
        void                    *port_argp)
 {
-       int                     port;
+       long                            port;
        int                     error;
        struct _ics_rinfo       *valptr = NULL;
        struct sockaddr_in      sin;
        struct socket           *sock, *newsock;
        mm_segment_t            oldfs;
 
-       port = (int) port_argp;
+       port = (long) port_argp;
 
        error = sock_create(PF_INET, SOCK_STREAM,
IPPROTO_TCP, &sock);
        if (error < 0)
@@ -1684,7 +1684,11 @@
        ASSERT(ptr != NULL);
        msg->msg_iov->iov_len =
(__kernel_size_t)(sizeof(clusternode_t)+sizeof(int));
 
+#ifdef __alpha__
+       ASSERT(sizeof(clusternode_t) == 8);
+#else
        ASSERT(sizeof(clusternode_t) == 4);
+#endif 
        /*
         * Now get the node number and channel number
from the client.
         * The node number is given in 4 bytes (not
int, not long...).
diff -Nru linux-2.4.2/cluster/ics/ics_tcp/ics_llsvr.c
linux/cluster/ics/ics_tcp/ics_llsvr.c
--- linux-2.4.2/cluster/ics/ics_tcp/ics_llsvr.c Fri
Jun 22 10:43:55 2001
+++ linux/cluster/ics/ics_tcp/ics_llsvr.c       Fri
Jun 22 11:22:56 2001
@@ -573,7 +573,8 @@
 
        if (nelem > 0) {
                /* Assert word alignment */
-               ASSERT((u32)param % sizeof(int64) ==
0);
+/* param is an address ane...@di... */
+               ASSERT((u_long)param % sizeof(int64)
== 0);
 
                lenleft =
ics_skb_bcopy(&handle->sh_msg_inline_skb_p,
                                       
&handle->sh_msg_inline_data_p,
@@ -646,7 +647,8 @@
 
        if (nelem > 0) {
                /* Assert word alignment */
-               ASSERT((u_int)param % sizeof(int32) ==
0);
+/* param is an address ane...@di... */
+               ASSERT((u_long)param % sizeof(int32)
== 0);
 
                lenleft =
ics_skb_bcopy(&handle->sh_msg_inline_skb_p,
                                       
&handle->sh_msg_inline_data_p,
diff -Nru linux-2.4.2/cluster/util/nsc_scalls.c
linux/cluster/util/nsc_scalls.c
--- linux-2.4.2/cluster/util/nsc_scalls.c       Fri
Jun 22 10:43:55 2001
+++ linux/cluster/util/nsc_scalls.c     Fri Jun 22
11:25:06 2001
@@ -543,6 +543,9 @@
                break;
        default:                                /*
invalid version */
                error = -EINVAL;
+#ifdef DEBUG
+               printk(KERN_DEBUG "ssisys: Invalid
Vesrion\n");
+#endif
                goto out;
        }
 
@@ -1539,7 +1542,7 @@
  *     ulong                   - as defined in
cluster.h
  */
 ulong
-cluster_get_api_state(ulong clms_state)
+cluster_get_api_state(ulong   clms_state)
 {
        switch(clms_state) {
                case CLMS_STATE_NEVERUP:
@@ -1560,7 +1563,7 @@
                        return CLUSTERNODE_DOWN;
                default:
                        /* invalid state */
-                       return 0;
+                       return 0x0;
        }
 }
 
@@ -1717,6 +1720,9 @@
        if (uaddrin_len <= 0 ||
            uaddrin_len > sizeof(inargs) ||
            uaddrout_len == 0) {                /* no
out arg */
+#ifdef  DEBUG 
+               printk(" Error in uaadrin_len or
uaddrout_len\n");
+#endif /* DEBUG */
                return -EINVAL;
        }
        if (copy_from_user((caddr_t)&inargs, uaddrin,
uaddrin_len)) {
@@ -1965,6 +1971,7 @@
        u_int                   uaddrout_len,
        rval_t*                 rvp)
 {
+
 
        return NSC_MAX_NODE_VALUE;
 }
diff -Nru linux-2.4.2/cluster/util/xdr_msghdr.c
linux/cluster/util/xdr_msghdr.c
--- linux-2.4.2/cluster/util/xdr_msghdr.c       Fri
Jun 22 10:43:55 2001
+++ linux/cluster/util/xdr_msghdr.c     Fri Jun 22
11:29:41 2001
@@ -32,6 +32,11 @@
 
 /*
  * Modification History:
+ * 22jun01 ane...@di...
+ *     xdrmsghdr_getpos()
+ *     xdrmsghdr_setpos()
+ *     Removed  long - int  casting issues while
finding the 
+ *     difference  between two address  
  *
  * 01apr01     ksung
  * - ported to use Linux msghdr structs.
@@ -42,6 +47,8 @@
  *   to handle that circumstance.  Also fixed
xdrmblk_putlong() in a
  *   corresponding way.  Manifested as errors in
osmlog of:
  *   "xdr_mblk: long crosses mblks!".
+ *
+ *
  */
 
 /* Avoid name collisions with <linux/sunrpc> code. */
@@ -350,8 +357,8 @@
 u_int
 xdrmsghdr_getpos(const XDR *xdrs)
 {
-       return ((u_int)xdrs->x_private - 
-               (u_int)((struct iovec
*)(xdrs->x_base))->iov_base);
+       return ((u_int)((u_long)xdrs->x_private - 
+               (u_long)((struct iovec
*)(xdrs->x_base))->iov_base));
 }
 
 /*
@@ -385,11 +392,11 @@
         */
        char *lastaddr = (char *)(xdrs->x_private +
xdrs->x_handy);
 
-       if ((int)newaddr > (int)lastaddr)
+       if (newaddr > lastaddr)
                return (FALSE);
 
        xdrs->x_private = newaddr;
-       xdrs->x_handy = (int)lastaddr - (int)newaddr;
+       xdrs->x_handy = lastaddr - newaddr;
 
        /*
         * Maintain coherency between the x_handy
buffer size and
diff -Nru linux-2.4.2/include/asm-alpha/signal.h
linux/include/asm-alpha/signal.h
--- linux-2.4.2/include/asm-alpha/signal.h      Thu
Jun 25 03:00:11 1998
+++ linux/include/asm-alpha/signal.h    Fri Jun 22
11:32:21 2001
@@ -64,13 +64,18 @@
 #define SIGINFO                29
 #define SIGUSR1                30
 #define SIGUSR2                31
+#define SIGMIGRATE     32
+#define SIGCLUSTER     33
+/* Added SIGUNUSED to make sure the signal numbers
remain same across 
+platform ane...@di... */
+#define SIGUNUSED      34
 
 #define SIGPOLL        SIGIO
 #define SIGPWR SIGINFO
 #define SIGIOT SIGABRT
 
 /* These should not be considered constants from
userland.  */
-#define SIGRTMIN       32
+#define SIGRTMIN       35
 #define SIGRTMAX       (_NSIG-1)
 
 /*
diff -Nru linux-2.4.2/include/asm-alpha/unistd.h
linux/include/asm-alpha/unistd.h
--- linux-2.4.2/include/asm-alpha/unistd.h      Thu
Jan 25 04:46:23 2001
+++ linux/include/asm-alpha/unistd.h    Fri Jun 22
11:32:59 2001
@@ -315,6 +315,9 @@
 #define __NR_mincore                   375
 #define __NR_pciconfig_iobase          376
 #define __NR_getdents64                        377
+/* BEGIN SSI SYSTEM CALLS */
+#define __NR_ssisys                    378
+
 
 #if defined(__GNUC__)
 
diff -Nru
linux-2.4.2/include/cluster/ics/ics_tcp_sock.h
linux/include/cluster/ics/ics_tcp_sock.h
--- linux-2.4.2/include/cluster/ics/ics_tcp_sock.h    
 Fri Jun 22 10:43:56 2001
+++ linux/include/cluster/ics/ics_tcp_sock.h    Fri
Jun 22 11:37:44 2001
@@ -18,6 +18,11 @@
  */
 /*
  * Header file used internally by the NSC ICS
low-level code.
+ *
+ * 22JUN2001  ane...@di...
+ * Change the ihdr_transid  field of struct icshdr_t.

+ * This variable is found to carry address
+ *
  */
 
 #ifndef        _CLUSTER_ICS_ICS_TCP_SOCK_H
@@ -38,7 +43,11 @@
        unsigned int    ihdr_service:32;
        unsigned int    ihdr_procnum:16;
        unsigned int    ihdr_inline_len:16;
-       unsigned int    ihdr_transid:32;
+#ifdef __alpha__
+       unsigned long   ihdr_transid:64;
+#else
+        unsigned int   ihdr_transid:32;
+#endif 
        unsigned int    ihdr_orignode:32;
        unsigned int    ihdr_chan:16;
        unsigned int    ihdr_unused:16;
diff -Nru linux-2.4.2/include/cluster/ics.h
linux/include/cluster/ics.h
--- linux-2.4.2/include/cluster/ics.h   Fri Jun 22
10:43:56 2001
+++ linux/include/cluster/ics.h Fri Jun 22 11:40:05
2001
@@ -14,6 +14,12 @@
  *     details.
  *
  *     Questions/Comments/Bugfixes to
ci-...@li...
+ */
+
+ /*
+ * 22JUN2001 ane...@di...
+ * Some int32 and int64 changes specific to alpha
platform 
+ *
  *
  */
 #ifndef _CLUSTER_ICS_H
@@ -29,13 +35,21 @@
  */
 #ifndef _INT32
 #define _INT32
+#ifdef __alpha__
+typedef        int int32;
+#else
 typedef        long int32;
+#endif 
 #endif
 
 
 #ifndef _INT64
 #define _INT64
+#ifdef __alpha__
+typedef long int64;
+#else
 typedef dl_t int64;
+#endif 
 #endif
 
 #define INT32MASK  3
diff -Nru linux-2.4.2/include/cluster/icsgen.h
linux/include/cluster/icsgen.h
--- linux-2.4.2/include/cluster/icsgen.h        Fri
Jun 22 10:43:56 2001
+++ linux/include/cluster/icsgen.h      Fri Jun 22
11:42:07 2001
@@ -16,14 +16,21 @@
  *     Questions/Comments/Bugfixes to
ci-...@li...
  *
  */
+/*
+ *
+ * 22JUN2001 ane...@di...
+ * Some int long encoding changes wrt to alpha
platform 
+ *
+ *
+ */
 #ifndef        _CLUSTER_ICSGEN_H
 #define        _CLUSTER_ICSGEN_H
 
 typedef struct cli_opaque_handle {
-               int     dummy;
+               long    dummy;
        } cli_handle_t;
 typedef struct svr_opaque_handle {
-               int     dummy;
+               long    dummy;
        } svr_handle_t;
 
 #include <linux/types.h>
@@ -310,6 +317,20 @@
  * param IN  u_long
  * param OUT u_long *
  */
+#ifdef __alpha__ 
+#define cli_encode_u_long(hndl,x) \
+       cli_encode_as_llong(hndl,x,u_long)
+#define cli_encode_u_long_p(hndl,x) \
+       cli_encode_as_llong_p(hndl,x,u_long)
+#define cli_decode_u_long_p(hndl,x) \
+       cli_decode_as_llong_p(hndl,x,u_long)
+#define svr_decode_u_long(hndl,x) \
+       svr_decode_as_llong(hndl,x,u_long)
+#define svr_decode_u_long_p(hndl,x) \
+       svr_decode_as_llong_p(hndl,x,u_long)
+#define svr_encode_u_long_p(hndl,x) \
+       svr_encode_as_llong_p(hndl,x,u_long)
+#else
 #define cli_encode_u_long(hndl,x) \
        cli_encode_as_int(hndl,x,u_long)
 #define cli_encode_u_long_p(hndl,x) \
@@ -322,6 +343,7 @@
        svr_decode_as_int_p(hndl,x,u_long)
 #define svr_encode_u_long_p(hndl,x) \
        svr_encode_as_int_p(hndl,x,u_long)
+#endif 
 
 /*
  * param IN  u_int
@@ -398,18 +420,33 @@
 /*
  * param IN clusternode_t
  */
+#ifdef __alpha__
+#define cli_encode_clusternode_t(hndl,x) \
+       cli_encode_as_llong(hndl,x,clusternode_t)
+#define svr_decode_clusternode_t(hndl,x) \
+       svr_decode_as_llong(hndl,x,clusternode_t)
+#else
 #define cli_encode_clusternode_t(hndl,x) \
        cli_encode_as_int(hndl,x,clusternode_t)
 #define svr_decode_clusternode_t(hndl,x) \
        svr_decode_as_int(hndl,x,clusternode_t)
+#endif 
 
 /*
  * param OUT clusternode_t *
  */
+#ifdef __alpha__ 
+#define cli_decode_clusternode_t_p(hndl,x) \
+       cli_decode_as_llong_p(hndl,x,clusternode_t)
+#define svr_encode_clusternode_t_p(hndl,x) \
+       svr_encode_as_llong_p(hndl,x,clusternode_t)
+#else
 #define cli_decode_clusternode_t_p(hndl,x) \
        cli_decode_as_int_p(hndl,x,clusternode_t)
 #define svr_encode_clusternode_t_p(hndl,x) \
        svr_encode_as_int_p(hndl,x,clusternode_t)
+
+#endif 
 
 /*
  * param IN:VAR:OOL clusternode_t *


____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

[CI] (no subject)

[<kva...@ya...>]

Hi , 

 Some minor changes needed to make cluster tools work
on Alpha  Linux. 
 
-aneesh 

____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

[CI] [PATCH]Cluster Infrastructure for Linux -Alpha Platform (SSI cluster)

[<kva...@ya...>]

Hi ,

Sending below the patch  that make Cluster
Infrastructure  work for Alpha Linux. Every thing
works fine with two nodes. I haven't verified it for
more than two nodes. But  I think  it  should work
fine even for more than two nodes.

Steps involved in applying the patch
1) Download kernel version 2.4.2 from alphalinux.org
2) apply the Cluster infrastructure  patch from
http://opensource.compaq.com/sourceforge/project/?group_id=22
3) Apply the following patch

If there is any issue please let me know.

 -aneesh



____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

[CI] STATUS OF ALpha PORT

[<kva...@ya...>]

Hi , 

 Today I was trying to add a second node, but  still
the kernel is crashing. Both the nodes crash and dumps
at the same point saying that unable to handle to
kernel paging at some address. Today I found many
places where 'int' to 'long' conversion issues were
there. Files like icsgen.h are  really dangerous :) 

I will try to fix it. Just thaught I will update the
status 

 -aneesh 

____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie

[CI] test

[Kai-Min S. <ka...@CS...>]

This is a test to see if outsiders can post to this mailing list.

-Kai

[CI] Re: [Dlm-devel] Compaq Cluster Manager & DLM

[David B. Z. <dav...@co...>]

Sorry for the long delay.  We have our open source web page working.  You can find
the CLMS/ICS sourceforge style set-up at http://opensource.compaq.com under
"Cluster Infrastructure for Linux."  I've been trying to find good design specs,
but they all need editing first.

Just this week we moved the initialization of CLMS/ICS out of the early kernel
initialization, and into system calls.  Now we can initialize the cluster from
linuxrc of the ramdisk or much later for loosely coupled clusters.  I'm going to
try to combine our initialization with GFS mounting of a shared root with external
lock manager from the ramdisk.  I assume that mapping still doesn't work, so I'll
have to come up to single-user.

If you have any questions about what you find on our web site, be the first to
send to the mailing list.    We are officially announcing our web page(s) next
week.

Peter Badovinatz wrote:

> --- "David B. Zafman" <dav...@co...> wrote:
> >
> > My group is about to open source our Cluster Manager code.  I would like to
> > discuss with you guys the prospects of utilizing it for your DLM project.  We
> > believe that is has the features which are relied upon from its HACMP roots.
> > I
> > haven't seen much mail this month, so I was worried that lots of development
> > was
> > occuring based on working around the lack of a real cluster manager.
>
> David,
>
> We're quite interested in working with you.  I have meant to contact you, but
> we've been in some amount of flux.  We have been violating the release early
> and often rule :(  In any case, we have not yet done lots of development
> directly related to lack of a cluster manager although we have done a fair
> amount of worrying and some head scratching.  More worringly, we've had some
> basic structural issues (internal state machine) that we needed to clean up.
> Actually, that Ian is cleaning up, as I've been yo-yoing among many different
> things.  Lab transitions, issues with our test rack (did you know the 2.4
> kernel will NOT load from CD onto a Netfinity 5500?  And that is our test
> cluster!)
>
> Ok, anyway, this sounds like great news. We've attempted to restructure the DLM
> code so that we can 'plug into' different cluster managers.  We've successfully
> done two so far, Heartbeat and IBM's (still proprietary) RSCT.  The latter
> isn't really a useful option for us in most situations because it's closed
> (although, like Compaq's code, it provides a "real" cluster manager.)
>
> We're hoping to get the wheels driving again, you should see some API spec
> updates from me being posted here soon (nothing really tied to cluster manager
> issues, but, necessary for users like GFS), and Ian should have his updates out
> ASAP if the lab machines are back in business.
>
> I am in Sydney, Australia until 25 June, but am otherwise attempting to work
> 'normally.'  Such as is normal for me anyway...  If you can point me docs (for
> now, code when it's available) that describe the APIs (or the semantics) the
> DLM would use to hook in, I can see how it fits into the framework we have so
> far.  I am not closely familiar with your code, but from what I do know, I
> think it will do us quite well!  Do you have a specific time frame for
> availability?
>
> Thanks for the great news!  We'll add it to the arsenal.
>
> >
> > --
> > David B. Zafman                             | Compaq Computer Corporation
> > mailto:dav...@co... | http://www.compaq.com
> > "Thus spake the master programmer:  When you have learned to snatch
> > the error code from the trap frame, it will be time for you to leave."
> >
>
> Peter
>
> =====
> These have been the opinions of:
> Peter R. Badovinatz -- (503)578-5530 (TL 775)
> wo...@us.../tab...@ya...
> and in no way should be construed as official opinion of
> IBM, Corp.
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail - only $35
> a year!  http://personal.mail.yahoo.com/

--
David B. Zafman                             | Compaq Computer Corporation
mailto:dav...@co... | http://www.compaq.com
"Thus spake the master programmer:  When you have learned to snatch
the error code from the trap frame, it will be time for you to leave."

[CI] Testing

[Brian J. W. <Bri...@co...>]

Nothing but a test.

-- 
Brian Watson            |  "The common people of England... so jealous
Linux Kernel Developer  |  of their liberty, but like the common
Compaq Computer Corp    |  people of most other countries never
Los Angeles, CA         |  rightly considering wherein it consists..."
310-414-3119            |      -Adam Smith, Wealth of Nations, 1776